# Moonlit Legal Research MCP

**Ask a legal question, get an answer built from the official publication, and check every citation at the source in one click.**

Connect Moonlit and your assistant can pull the judgments interpreting a specific article, compare a rule across countries in one prompt, retrieve the official text of any provision, and draft memos where every claim of law carries a verified footnote. Every answer is built from documents retrieved live from Moonlit's Data Layer, each carrying a direct URL to the original publication, and a citation graph links each provision to the case law and guidance that cite it. The same Data Layer runs the research of the Dutch Supreme Court, the Public Prosecution Service, the Dutch Tax Administration and the Dutch Central Bank. Deloitte, KPMG and legal AI companies such as Eudia and Saga build on it, alongside universities, law firms and in-house legal teams.

## Moonlit's Data Layer

| | |
| --- | --- |
| Jurisdictions | 35+ |
| Official portals | 130+ |
| Sources (courts, legislatures, regulators) | 3,900+ |
| Documents | 20M+ |
| Citation references | 100M+ |

Europe first: the EU institutions, the ECHR, and national law across the Netherlands, Germany, France, Belgium, Luxembourg, Austria, Switzerland, Italy, Spain, Portugal, Poland, the Nordics, Central Europe, Ireland and the UK, with international sources such as the UN and the OECD. Global expansion is underway: Brazil, Australia, South Korea, Japan, Hong Kong and the Gulf are live today, and further jurisdictions are added on customer request. Ingestion runs daily against the official portals and legal experts review and enrich the data.

Alongside the primary law: insights from 100 international law firms and practitioner publishers, 350,000+ articles, so a research run can set what practitioners say next to what the sources hold. This collection is searched by keyword. Full detail at [moonlit.ai/coverage](https://www.moonlit.ai/coverage).

## Tools

Nine read-only tools.

| Tool | What it does |
| --- | --- |
| `keyword_search` | Exact words, phrases and Boolean expressions. Best when the ECLI, CELEX or article number is known. |
| `hybrid_search` | Natural-language question, keyword plus semantic search fused into one ranked list. |
| `hybrid_search_reranked` | Hybrid search with a reranking pass. The most precise on complex questions. |
| `reference_search` | Documents that cite a given regulation or judgment, or that it cites. Works at article level, so it answers "which judgments interpret Article 22" instead of "which documents mention the GDPR". |
| `get_filters` | Canonical lists of jurisdictions, portals, document types, fields of law and sources. |
| `retrieve_document` | Full text and metadata of a single document, including the source URL at the publisher. |
| `get_document_articles` | Article-level structure of a composite document such as the GDPR or the AI Act. |
| `convert_to_celex` | Directive or regulation number to its canonical CELEX identifier. Always free. |
| `get_account_status` | Your plan and remaining monthly quota. Always free. |

## Connect in five minutes

```
https://mcp.moonlit.ai/mcp
```

You sign in with a Moonlit account; no key is needed and the free tier requires no card.

1. Create an account at [app.moonlit.ai](https://app.moonlit.ai/).
2. In Claude, go to **Settings**, then **Connectors**, click **Add custom connector**, name it **Moonlit** and paste the server URL.
3. Click **Add**, sign in and approve the connection. Tools are available in your next chat.

First prompt: "Find the official text of Article 6 of the EU AI Act." The response cites CELEX 32024R1689 with the source URL at the publisher, followed by the article text.

The server works in every client that supports remote MCP: ChatGPT, Microsoft 365 Copilot, Langdock, Le Chat, Perplexity, and developer tools such as Claude Code, Cursor and VS Code. Per-client setup: [How to connect](https://www.moonlit.ai/docs/mcp/connect). Organizations can pool usage and create MCP keys for deployed services and agents: [Organizations](https://www.moonlit.ai/docs/mcp/organizations).

## Example prompts

The same tools carry cross-border research, regulatory monitoring, multi-jurisdictional due diligence and multilingual retrieval pipelines. Each prompt below shows a different capability; more at [Example prompts](https://www.moonlit.ai/docs/mcp/example-prompts).

**Multilingual comparison.** Compare the enforceability of non-compete clauses in employment contracts across the Netherlands, Germany and Belgium. Surface the key recent decisions and the controlling statutory provisions. The assistant queries each jurisdiction in its national language, so one English prompt reaches Dutch, German and French sources.

**Citation graph.** Show me Article 22 of the GDPR and the leading EU and Dutch judgments interpreting it since 2020. One-line holding for each.

**EU law into national law.** Draft a one-page memo on a VAT exemption under Article 132 of the VAT Directive: the Directive text, the national implementation and the tax authority's guidance, each cited at the source.

## Skills

Three first-party skills teach an AI assistant to work with Moonlit the way a careful lawyer would: governing law first, verified at the source, and no claim of law without a citation that can be checked in place. Reviewing an answer becomes a matter of opening the citations. Install takes one upload; from then on the skill applies automatically.

| Skill | What it governs | Install it when |
| --- | --- | --- |
| [moonlit-operator](https://www.moonlit.ai/docs/mcp/skills/moonlit-operator) | How sources are found, verified and cited | Any legal work with Moonlit |
| [moonlit-reasoning](https://www.moonlit.ai/docs/mcp/skills/moonlit-reasoning) | How law is applied to facts once sources are established | The assistant should analyze the law, beyond reporting it |
| [moonlit-memo](https://www.moonlit.ai/docs/mcp/skills/moonlit-memo) | The full research memo, both disciplines built in | You want research memos; it works alone |

The skill files are versioned in the open at [github.com/moonlit-ai/moonlit-skills](https://github.com/moonlit-ai/moonlit-skills) (MIT), with per-app install guides at [Skills](https://www.moonlit.ai/docs/mcp/skills). They work in claude.ai, ChatGPT, Microsoft 365 Copilot, Langdock, Mistral and Claude Code. Building your own verification-first skill? Moonlit's tools are a retrieval layer to ground it on; the [Data API](https://www.moonlit.ai/docs/api) documents the full search surface.

## Plans

Searches, document retrievals and filter lookups count as one call each. `get_account_status` and `convert_to_celex` never count. Prices excl. VAT

| Plan | Calls per month | Price |
| --- | --- | --- |
| MCP Free | 100 | EUR 0, no card required |
| MCP Basic | 500 | EUR 39 per user |
| MCP Pro | 1,200 | EUR 79 per user |
| Enterprise | Above 1,200, pooled organization usage | [Book a call](https://www.moonlit.ai/book-a-call) |

Ask your assistant for your Moonlit account status any time to see your remaining quota.

## Security and data handling

Moonlit Legal Technologies B.V. (Amsterdam) is ISO 27001 certified and GDPR compliant, with EU-hosted infrastructure on Microsoft Azure. Requests are logged at the request level; the content of queries is never stored and never used to train models. Legal framework: [MCP Terms](https://www.moonlit.ai/docs/legal/terms-mcp), [Privacy Policy](https://www.moonlit.ai/docs/legal/privacy-mcp), [DPA](https://www.moonlit.ai/docs/legal/dpa), [Sub-processors](https://www.moonlit.ai/docs/legal/sub-processors), [Trust Center](https://trust.moonlit.ai/).

Outputs are research material with verifiable sources; they do not constitute legal advice.

## Support

Documentation at [moonlit.ai/docs](https://www.moonlit.ai/docs), version history in the [changelog](https://www.moonlit.ai/docs/changelog), support at [support@moonlit.ai](mailto:support@moonlit.ai).